Visual Odometry to estimate the distance to a known object.

Visual odometry
Name
Institution
EMAIL
ABSTRACT
This project studies the use of visual odometry to locate the position of a known object. The object is located by the use of HSV. HSV, standing for hue, saturation and value is method used to represent points in the cylindrical coordinate using RGB color model. IN HSV, the RGB geometry is rearranged in order to make it more perceptive than the Cartesian representation. Location of the Threshold was done by changing the values of the HSV [1]. After this a structuring element was used to dilate and erode the image. This makes it possible to determine the image size. Knowledge of the size of the object in the image makes it possible to determine its distance by use of a mathematical function.
* INTRODUCTION
Vision odometry is the procedure used to estimate an agent`s egomotion. The agent may be a human being, vehicle or even robot. Procedure employs the use of a single or multiple cameras attached to the body of the agent. Visual odometry is employed in the field of robotics, automotive industry and in wearable computing. Owing to its close resemblance to wheel odometry, this procedure was termed as visual odometry courtesy of Nister`s landmark paper of 2004 [3]. The working principle of wheel odometry is based on estimation of a vehicles motion in augmentation via integration of the wheels turns over the time taken. On the other hand, Visual odometry works by evaluation of the vehicle position by analyzing the changes caused by motion on the onboard cameras images. Visual odometry requires adequate radiance in the environment. In addition the presence of stationary scene with substantial coarseness is required which will allow the extraction of the apparent motion. In order to attain adequate scene overlap, the frames need to be captured in a successive manner.
* OBJECTIVE
* To locate an object and get its center in the x-y plane using visual odometry.
* To estimate the distance to the object using visual odometry.
* METHOD
* Materials used
* HD camera logittech c920 1280*720
* Opencv library code written in C++
HSV was used to locate the object by changing the values of the HSV in order to get the threshold. This is as indicated below:
H min =0
Hmax=256
S min =0
S max = 256
V min =0
V max =256
The object color is red. The value of the s min was varied to 148 and a white pixel is obtained. This represents the object in the threshold. A structuring element was then created to be used to dilate and erode the image. An image element of 3px by 3px rectangle was chosen. A circle was then created in the middle of the original image of the object located in the threshold. This was used to identify the location in the x-y plane (height, width). The white pixel from the threshold was then summed up. Having an idea of the size of the object in the image, the distance to the object was obtained by use of the following function:
distanc...